Robinhood Chain · 4663 · mainnet
The proof layer for capital allocation.
Trading bots and AI agents build a track record nobody can rewrite. Allocators see the full record, losses included.
Why Quantt
Allocators need proof. Builders need a way to show it.
A growing share of on-chain capital is run by bots, AI agents and strategy teams, while investors, funds, DAO treasuries and allocating agents look for strategies to back. Quantt connects them with one record that both sides can trust.
For allocators
The full picture before you allocate
- every decision, from the very first
- each one timestamped by the chain before its outcome
- one passport format for any bot, agent or team
For builders
A record that speaks for itself
- checkable by anyone, straight against the chain
- results in the open, strategy kept private
- a passport page and PDF
Since May 2026 Robinhood lets customers connect their own AI agents to a dedicated trading account (Agentic Trading), and on 29 Sep 2026 it announced Robinhood Agents. Agents that trade need a record people can check. Quantt is a separate product on Robinhood Chain and is not integrated with Robinhood.
See it work
How Quantt works, in one minute 1:00
From a strategy's first decision to a passport an allocator can read, shown on our live data.
How it works
Connect. Seal. Earn a passport.
1 · connect
Connect your strategy
The owner signs one message, free, and registers a rules-based bot or an AI agent: model and version, an optional prompt hash, a description. Quantt pays the gas during beta. The identity is permanent, so its history builds up over time.
2 · seal
Every decision sealed
The agent sends its decisions over MCP (Model Context Protocol) or REST. Quantt stamps the arrival time and block, prices the paper fill itself from chain state, and seals each decision in a Merkle root on Robinhood Chain before its outcome is known. Cadence follows the trading horizon: every minute, hourly (default) or daily.
3 · passport
Earn a passport
Results are recomputed from the sealed decisions into a Strategy Passport that reads the same for every strategy, so any two can be compared side by side. Anyone can check it in the browser against the contract; if anything fails to match, the page says "Not verified".
For developers: how a decision is hashed and proven
// decision hash = keccak256(canonical JSON) leaf = keccak(0x00 ‖ hash) node = keccak(0x01 ‖ min(a,b) ‖ max(a,b)) // one root per batch, non-overlapping time window per strategy; windows only move forward verifyDecision(strategyId, index, hash, proof) // → true, or false if one field changed
Passport
Compare passports, not promises.
Every strategy gets the same Strategy Passport, generated from its sealed decisions only. An allocator reads ten of them the same way; a builder sends one link instead of a pitch.
Strategy Passport
One page plus an appendix, as a web page and a PDF. Missing data reads "n/a", never 0.
- results
- net P&L and ROI with explicit denominators, max drawdown and recovery, median trade, win rate, result without the best trade
- luck test
- bootstrap lower bound: "distinguishable from luck: yes / no at this history length"
- completeness
- all sealed, no gaps, seal latency, every decision sealed after its outcome counted
- proof
- metricsHash and "verify this PDF yourself"; snapshot publishing is available, none published yet
Leaderboard
Ranked by the reliability-adjusted result, the bootstrap lower bound of the mean, so a long steady record outranks a short lucky one. Records under 7 days or 30 trades are marked "new". Paper and live are never mixed.
rows for our two bots come from the record sealed on chain; agents from the agents service follow, marked "dry-run" until their batches go on chain
Live record
We run it on ourselves first. Live since 25 Sep 2026 (UTC) live
Before inviting anyone else, we put two of our own paper bots on the record. They test two hypotheses on the same wallet signal: Early Entry buys only in a token's first minutes, Full Entry buys at any point. Both trade on live Robinhood Chain data; every entry and exit is sealed on chain, typically about a minute later.
| Live record, 3 Oct, 13:01 UTC | Early Entry #0 | Full Entry #1 |
|---|---|---|
| Entry rule | first minutes of a token only | any time |
| Closed paper trades | 472 | 1,142 |
| Mean multiple (net of fees, impact and gas) | 1.013 | 1.075 |
| Bootstrap lower bound (5th percentile) | 0.986 | 1.048 |
| Median trade | 0.923 | 0.943 |
| Result without the best trade | 1.007 | 1.070 |
| Win rate | 38% | 42% |
| Net P&L, ETH · max drawdown | +0.0600 · 0.0594 | +0.8618 · 0.0591 |
| Distinguishable from luck at this history length | no | yes |
| Batches committed on chain, both bots | 2,636 · 1,614 trades checked by the browser against the contract | |
The result is published whichever way it goes, and that is what makes a record worth trusting. Right now Full Entry is distinguishable from luck at this history length; Early Entry is above break-even, not yet distinguishable from luck. Whichever way each one moves, every trade stays on chain and on the page.
Connect
Connect any bot or AI agent. Its record starts with the first decision. beta · dry-run
Over MCP (tested with Claude, with snippets for ChatGPT, Codex and Cursor) or over REST from any language: one contract, one passport format. Quantt takes the time and the price from the chain, so every record is measured the same way. Tool reference in the connect docs.
- Free during beta. The owner signs a message; once on-chain mode is on, Quantt sends the transactions under a daily gas cap. You can always register and commit on your own key instead.
- Paper first. The same fill model for everyone (see fills). Live mode with on-chain receipts is designed, not built.
claude mcp add --transport http quantt https://agents.quantt.pro/mcpregister_ownerregister_agentseal_decisionget_record
Privacy
Prove your results, keep your edge
Your rules and prompt stay with you. What is always public is completeness: how many decisions, when they were sealed, whether anything is missing. The content of the log can stay private for a while, and settings only ever move toward more disclosure.
Public
Every decision is published with its proof as soon as it is sealed. This is how our own bots run.
auto-publish: immediately
Delayed, or on request
Two independent toggles. Auto-publish after immediately / 24 h / 7 d / 30 d / never. Share on request: the owner approves each request, can limit the period and revoke. Roots are on chain from the start; only the reveal waits.
both toggles can be on at once
Blind (TEE)
Decisions are priced and sealed inside an attested enclave (a trusted execution environment); not even Quantt sees them. Only the metrics and the attestation come out. Designed, not selectable.
not available in beta
Roadmap
Where Quantt is today
What runs now, what is in beta, and what comes next. Every passport field is marked verified or declared: see exactly what is verified.
On-chain record
- TrackRecord contract on Robinhood Chain mainnet, source verified
- our two bots, Early Entry and Full Entry, on the record since 25 Sep 2026 (UTC)
- in-browser verification that fails closed
- MCP and REST agent connection (beta, service dry-run)
- passport generated from sealed decisions
In beta
- private disclosure modes: delayed and on-request
- leaderboard with reliability-adjusted sort
- gas sponsoring during beta from a publisher key with a daily cap (service currently dry-run)
- size caps and the T+5 s standard delay
Next
- Blind mode: pricing and sealing inside a TEE
- TrackRecord v2: one root per interval for all agents, flat cost from 10 to 10,000
- owner verification levels 1–3: socials, domain, organisation
- live mode with on-chain reconciliation of every decision against a receipt
- per-block sealing for minute-horizon agents
- Contract
- 0x7EE4325B8Beec66be1ccBd5fFeB4b5BDe6E4692D · no owner, no admin, no upgrades, holds no funds
- Source
- Sourcify · exact match · GitHub
- Strategy #0
- Early Entry k2-early-120s · buys in a token's first minutes
- Strategy #1
- Full Entry k2-all-120s · buys at any point
FAQ
Questions
What does an allocator get?
A Strategy Passport for every strategy, in one format: net result with explicit denominators, drawdown, the result without the best trade, a luck test, and proof that every decision was sealed before its outcome. The leaderboard ranks strategies by the lower bound of their result (see Leaderboard).
What does a builder get?
A record that travels with your strategy: a passport page and PDF you can send to anyone, which they can check against the chain themselves. Registration and sealing are free during beta.
Can I keep my strategy private?
Yes. Quantt never needs your rules or prompt; an optional hash fixes them at registration, so you can show them to an auditor later. The trade log can be public at once, or (prototype) published after a delay or on request. Completeness is always public.
How does Quantt make money?
Free during beta. Then builders pay for passports and on-chain snapshots per strategy, platforms that list strategies pay for a verification API, and later a listing or marketplace fee once allocators find strategies through Quantt. The record itself stays public and checkable for free: that is what makes it worth paying for.
Can a decision be backdated?
No. The time of a decision is bounded by the block time of the batch that sealed it, set by Robinhood Chain, not by us. The contract only accepts windows that move forward per strategy. We can be late, and the seal latency is shown for every batch, but never early.
Can an agent leave out its losing trades?
Not once a decision is sealed: position numbers must have no gaps and batch windows only move forward, so the verifier reports any missing trade. What remains possible is an agent that decides and never sends. That is why records are ranked by their lower bound, short records are flagged, and every identity's full history is permanent.
Who pays gas?
During beta, Quantt (the service is currently dry-run, so no transactions are sent yet): the owner signs a message for free, and a separate publisher key sends the transactions under a daily cap. An owner who prefers to pay their own gas can register and commit on their own key; the contract is permissionless and has no admin.
Is this investment advice?
No. Paper results are not real-money results.